People fixate on blank spaces if visual stimuli previously occupied these regions of space. This so-called "looking at nothing" (LAN) phenomenon is said to be a part of information retrieval from internal memory representations, but the exact nature of the relationship between LAN and memory retrieval is unclear. While evidence exists for an influence of LAN on memory retrieval for visuospatial stimuli, evidence for verbal information is mixed. Here, we tested the relationship between LAN behavior and memory retrieval in an episodic retrieval task where verbal information was presented auditorily during encoding. When participants were allowed to gaze freely during subsequent memory retrieval, LAN occurred, and it was stronger for correct than for incorrect responses. When eye movements were manipulated during memory retrieval, retrieval performance was higher when participants fixated on the area associated with to-be-retrieved information than when fixating on another area. Our results provide evidence for a functional relationship between LAN and memory retrieval that extends to verbal information.

Recent research suggests that when people retrieve information from memory they tend to fixate on the location where the information had appeared during encoding. We used this phenomenon to investigate if different information is activated in memory when people use a rule- versus a similarity-based decision strategy. In two studies, participants first memorized multiple pieces of information about various job candidates (exemplars). In subsequent test trials they judged the suitability of new candidates that varied in their similarity to the previously learned exemplars. Results show that when using similarity, but not when using a rule, participants fixated longer on the previous location of exemplars that resembled the new candidates than on the location of dissimilar exemplars. This suggests that people using similarity retrieve previously learned exemplars, whereas people using a rule do not. The study illustrates that eye movements can provide new insights into the memory processes underlying decision making.

When trying to remember verbal information from memory, people look at spatial locations that have been associated with visual stimuli during encoding, even when the visual stimuli are no longer present. It has been shown that such "eye movements to nothing" can influence retrieval performance for verbal information, but the mechanism underlying this functional relationship is unclear. More precisely, covert in comparison to overt shifts of attention could be sufficient to elicit the observed differences in retrieval performance. To test if covert shifts of attention explain the functional role of the looking-at-nothing phenomenon, we asked participants to remember verbal information that had been associated with a spatial location during an encoding phase. Additionally, during the retrieval phase, all participants solved an unrelated visual tracking task that appeared in either an associated (congruent) or an incongruent spatial location. Half the participants were instructed to look at the tracking task, half to shift their attention covertly (while keeping the eyes fixed). In two experiments, we found that memory retrieval depended on the location to which participants shifted their attention covertly. Thus, covert shifts of attention seem to be sufficient to cause differences in retrieval performance. The results extend the literature on the relationship between visuospatial attention, eye movements, and verbal memory retrieval and provide deep insights into the nature of the looking-at-nothing phenomenon.

In sequential diagnostic reasoning, observed pieces of evidence activate hypotheses in memory and are integrated to reach a final diagnosis. The order of evidence can influence diagnostic reasoning. This article examines the processing of ambiguous evidence underlying order effects if multiple hypotheses are activated. In five experiments with a quasi-medical scenario, participants dealt with symptom sequences supporting multiple diagnoses. The symptom order, the response mode (end-of-sequence, step-by-step), and the consistency of evidence were manipulated. A primacy order effect occurred with both response modes suggesting that ambiguous pieces of evidence were distorted toward the hypothesis that strongly corresponded with the first piece. The primacy effect was partially counteracted by stepwise belief ratings, which strengthened the weight of recent evidence and promoted switching to an alternative diagnosis. We conclude that once hypotheses are generated, the interplay of coherence-oriented information distortion and memory-dependent analytic processes propagates into distinct order effects in diagnoses.